This reminds me of a great interview with two ER doctors on Amanpour (PBS) two weeks ago. Pulse oximeters can be used to help detect which patients need hospitalization before they themselves notice, with astounding results.
> With the oximeter, Samaritan is asking patients to keep track of its measurements twice a day and to write the readings down. Then Home Health follows up with telehealth visits, either over the phone or via Zoom on days two and four, as well as days seven and 10.
This is happening now in hospitals with Covid-19. I know, because I'm married to a nurse. Nurses call them "happy hypoxics" because they have no idea their SPO2 levels are dangerously low -- but a pulse oximeter clearly indicates they need additional oxygen.
I had a pulmonary embolism‘s was white as a ghost struggling to breathe and that dang meter said everything was fine.
Even the bloodwork at the hospital said everything was fine.
Lost some lung over that and took a year to recover.
You might have been going into shock. A pulmonary embolism can block blood flow to part of a lung and kill that tissue but it won't necessarily cause SpO2 to drop much. Patients who are otherwise healthy have extra lung capacity to spare.
The current clinical recommendation for ambulatory COVID-19 patients at home is to go to the hospital if the pulse oximeter reading drops below 94%. Accurate pulse oximeters are cheap now so it's a good item to keep in your home first aid kit.
Mine frequently drops to the low 90s (or even 80s) based on the Apple watch, despite normally being 98-100. I'm certainly not in good shape, but still, should I be worried?
Wrist devices are generally not very accurate. I mean they can be accurate under ideal conditions, but if you're moving around a lot or sweating or the band is loose or you have dark or hairy skin then you might get a low reading. Better to use a real pulse oximeter that clips onto your finger. They're cheaper than an Apple Watch.
Well I don't think it sounds like Covid-19 specifically, but it wouldn't hurt to get that checked out. It's not something that happens to most people, if that's what you're asking.
Mine does as well. I attribute it to the fact that my arms aren't skinny, and I can impede blood flow by, e.g., "crooking" my elbow and compressing blood vessels there. I also sleep on my arm, which I suspect likewise impedes blood flow to the wrist area.
If I stand up, take a few deep breaths, and then rest my arm on a level surface I get high 90's right away. So i don't worry about the outliers (much).
This has been standard practice in India for a while now https://mumbai.apollohospitals.com/covid-19-update/stay-i-at... Package includes oximeter, test sample collection, PPE supplies and optional food drop off. Includes a zoom cal with nurse/doctor at regular touchpoint intervals to facilitate hospitalisation
More than 2 doctors have reported positive effects with certain drugs to treat the infection, but those people are lambasted. What makes this anecdotal evidence any different? Shouldn't we report this as misinformation since it didn't come from an approved source such as the CDC or WHO?
So some big limitations, speaking as a doctor. Firstly, I'm surprised that they allowed watches on the ward. They're a huge infection risk and most places are bare below the elbows (NHS), moreso for covid. Secondly, they only looked at symptomatic covid cases (1 day before symptoms started was their prediction as far as i can tell. This is completely useless as the 24hrs before symptoms start the patient is likely to already be highly infectious, and misses the asymptomatic carriers which have an indeterminate effect on spread as well. Far better to use the semi-proven (but equally controversial) lateral flow tests and avoid these risks. In the UK, hundreds of healthcare staff have been identified as covid+ >48hrs before symptoms in this way.
Speaking as a non-doctor, I don't have any lateral flow test supplies or equipment. Do they sell that on Amazon?
But seriously, a test I don't have easy access to is not an alternative. A lateral flow test has no predictive value to me because I'm never going to get one before I'm actually sick.
> But seriously, a test I don't have easy access to is not an alternative.
Do you have access to the custom Apple Watch app they used here? I don't either.
Health authorities could make either of these available to the general public -- the parent poster is making a comment about which is more valuable under the assumption of access.
You're not wrong, a test you don't have access to obviously has no predictive value for you, but right now we (the public) have access to neither, so this isn't a distinguishing factor.
Now, if you're making the argument that the Apple Watch app would be easier to distribute, then you can read the parent's criticism as "sure, but only knowing after you've potentially spread it is not very useful."
> Do you have access to the custom Apple Watch app they used here? I don't either.
That's pretty immaterial when discussing theoretical effectiveness. Obviously, getting wide use depends upon it getting packaged up into an app listed on the app store, and some sort of awareness push.
That needs to be done once, by the app's authors. For lateral flow tests, to have similar effectiveness, that means that literally every Apple watch wearer (just to assume a similar population size) needs to get themselves a supply of tests (and keep getting new tests when they run out), and remember to perform the test every day. That is a much higher hurdle to jump than getting this app onto people's watches (assuming their study is correct and can be independently verified!). I also expect the supply of test kits is not robust enough to handle that kind of demand.
> Now, if you're making the argument that the Apple Watch app would be easier to distribute, then you can read the parent's criticism as "sure, but only knowing after you've potentially spread it is not very useful."
Reducing spread potential by 24 hours when you've perhaps been spreading it for a week already isn't exactly a game changer, but it's still useful, especially if the only thing people have to do is install an app on their watch. Expecting everyone to obtain rapid test kits and test themselves daily is in general a non-starter.
> That needs to be done once, by the app's authors.
Except...it also needs FDA approval as a diagnostic, etc., which is probably easier said than done.
Obviously it's easier to distribute an app than lateral flow tests -- but the latter are much more effective too.
Clearly, we should do both, and...
> Reducing spread potential by 24 hours when you've perhaps been spreading it for a week already isn't exactly a game changer, but it's still useful, especially if the only thing people have to do is install an app on their watch. Expecting everyone to obtain rapid test kits and test themselves daily is in general a non-starter.
This totally depends on your goals. For the individual with limited outside exposure, is knowing 24h before you feel symptoms useful? Absolutely.
For a hospital or other institution (school?), who can order in bulk, and then mandate, lateral flow tests every day before work starts and observe their correct use (without having to rely on self-reported app results), the tests seem much more useful.
The data is all readily available in the default Apple health app.
Yes, to make these same predictions you'd have to do some statistics as described in the paper, but no special app is required.
(And that's why this finding is interesting: they have the data available from patients before they were sick, because every new Apple Watch records this data out of the box.)
Yes, the finding is definitely interesting because it suggests that we could make this available very broadly (i.e., anyone with an appropriate Apple Watch) without needing to ship a bunch of hardware.
But the limitations are worth noting: you may already have infected people by the time this detects your illness. That isn't true for lateral flow tests, making the latter more useful as a preventative testing tool (again, in the absence of access differences).
However, I don't understand what you're trying to say with this claim:
> Yes, to make these same predictions you'd have to do some statistics as described in the paper, but no special app is required.
"Do some statistics" is approximately as impossible as "order lateral flow test reagents" to the general public -- a consumable version of either the process described in this paper, or the lateral flow kits mentioned earlier, needs to be produced by someone skilled here, in order for this to be useful to the general public.
Getting a warning 24 hours "early" that you might be infected is still useful. Perhaps in that 24 hours you were going to make a grocery trip and potentially infect a bunch of people, but now you won't because you got that early warning.
Claiming that something is useless because it doesn't cover all cases seems a little weird to me. It's a tool in the toolbox that can reduce the possibility of spread in a limited set of cases.
Using lateral flow tests requires everyone to purchase a big supply of them and use one every day. This watch-based warning just requires you to have an Apple watch and run a new app on it (which I hope they'll make publicly available), and it'll just let you know without any extra action taken. Your average Apple watch user is probably pretty likely to take the simple step of installing a new app; they're not going to buy a supply of lateral flow tests (which they probably don't even know much about or haven't even heard of) and test themselves every day.
Yup. I was lucky the person whom I contracted COVID from informed me the moment they got the test result. I was on my way to work, just about to leave the car. Turned around and went back home. Developed symptoms the same evening.
It was a lucky coincidence that was Monday so I've stayed home the whole weekend and had no contacts for a couple of days I was likely already infectious. (The person I got it from only developed symptoms two days after our contact).
People who do contact tracking only on Wednesday.
So my contact informing me saved three days worth of my potential contacts from getting COVID from me.
Hope you're fine now.
Can I ask how did you contract it? What I mean is: did you and the other person follow the usual distancing rules (mask + 1 meter apart + no touching)?
I can only comment on the study in front of me, and not the use cases that 'might have been'. In this case, they tested in a healthcare setting as there is a presumed higher risk of catching covid amongst participants. Would the same false positive rate be true in the community, where transmission is much lower? We don't know because that wasn't the experiment.
The study authors claim that this is useful in the healthcare setting- I'm saying that it's useless because there are better tools in the arsenal (lateral flow tests), and watches are an infection risk.
Did they perform regular PCR testing on staff? No, and so they didn't detect any asymptomatic carriers. By comparison, all NHS staff are doing biweekly lateral flow tests, and hundreds of staff are detected prior to symptoms this way. This means that fewer staff are off sick/isolating, and more staff are available to care for patients.
It’s a study. A completely harmless, simple study that does not preclude anything else and shows positive results. What is your complaint? That it isn’t 100% definitive in all cases? That something else is better? That’s not the way progress is made.
Something that detects 10% of cases and is completely free (using a device people are wearing anyway) is valuable. You can be sure that some of these cases would have been missed by all the other tests, simply because not every one of these cases got any test at all.
> This is completely useless as the 24hrs before symptoms start the patient is likely to already be highly infectious, and misses the asymptomatic carriers which have an indeterminate effect on spread as well.
Nothing is a one-shot-solve-all solution to Covid. If Apple watches can detect it and prompt people to go get a full test, that’s great. I’ve got an ongoing theory that a lot of the spread is by people who ignore their symptoms because ‘they have a cough but it’s not a covid cough’ or ‘they haven’t lost their smell so it can’t be covid’ so the more things that can prompt you to get a test and isolate the better.
A "bare below the elbows" policy makes sense and there indeed are studies corroborating that pathogens can accumulate on wristwatches.
However, having worked as a paramedic I'm wondering how you would actually measure a patient's heart rate without a watch and with no additional equipment around.
Not every patient will be on cardiac monitoring, after all.
Yes, there are those nurse watches you can attach to your shirt or white coat but in order to take a proper look at those they have to be touched as well, which kind of defeats the purpose.
I assume it's because they accumulate crud and bacteria. You can wash your hands but you never really wash your watch. I also assume that because it's on your wrist every time you scratch your body or touch your face the bacteria covered watch deposits something icky on contact.
For reducing infection risk one option is to keep the watch in a pocket while working on the ward and wear a separate wireless heart rate monitor on the chest or upper arm. I don't think they did that in this study, just pointing it out as an option for healthcare workers who want to track their HRV on the job.
it is not unheard of for something at level x to be developed to level 2x so if it can do 24h now, with higher sample rate and better maths, what if it got to 48h in tractable time?
to the watch-in-a-ward point: valid. But, if this technique did get to 48h+ predictive quality, and was available to wearers of smart watches outside of wards would you still have hesitancy to consider it useful?
The utility of the test decreases with reduced risk. A 99% specific test is good when there are high levels of disease going around (e.g. covid in a hospital). In the community, rates are much lower, so the positive predictive value of the test is much lower, even with 99% specificity.
My conclusion is that if you tried this in a community setting, the number of false positives would far outweigh the true positives, and lead to dismissal of alerts.
"Longitudinally collected HRV metrics from a commonly worn commercial wearable device (Apple Watch) can identify the diagnosis of COVID-19 and COVID-19 related symptoms."
Having first-hand experience, I can confirm that my heart rate was WAY up, and my pulse-ox was WAY down when the virus was at its worst. I used the pulse ox value to make sure I was getting in good breaths (and yes, I know there is a hefty margin of error on these consumer devices.) I use a Garmin, but the sensors are pretty much the same on Apple Watch.
It would be pretty easy for the device to tell I was sick but probably impossible to tell I had COVID-19 instead of the Flu or a bad infection or maybe pneumonia.
In my case HRV went down almost 30% before I started to have symptoms with 50% of average on the worst day. Resting heart rate went up 35% on the worst day and respiratory rate went up 18%. But only HRV was obviously changed before I started to feel symptoms and became infectious.
umm speaking from personal experience (comparing garmin reading with hospital oximeter), garmin is waaaay inaccurate sometimes! Maybe it is because I always shower with it, or maybe it just loses accuracy over time. Didnt give it a second thought and just disabled the sensor.
Title is misleading IMO. The study has nothing specific to do with Apple Watch. They mentioned that they used "a commonly worn commercial wearable device (Apple Watch)" for testing. Any device with continuous heart rate monitoring will do.
> Your comment is supposition. Likely correct, but not guaranteed.
No, the paper was very clear that Apple Watch was used to measure Heart Rate Variability (HRV). There is absolutely no mention of any unique advantages that Apple Watch provided in the given scenario. The study is about using the HRV data from a wearable device to identify indicators of COVID-19; it is not an endorsement or an advertisement for Apple.
Have you compared the HRV data from different devices? Have you compared the Apple HRV data and the data from other devices to a reference instrument, or confirmed that it meets some standard?
The Apple Watch is the reference instrument for this paper. It would be a scientific error to conceal that fact.
This isn’t about product endorsements, it’s about saying what laboratory equipment was used so that other scientists don’t have hidden confounds when they attempt to replicate.
> The Apple Watch is the reference instrument for this paper.
Exactly. It is an instrument used to gather metrics. The study is about analyzing the metrics and not so much about the tools used to gather those metrics. Read the paper's title.
> Title is misleading IMO. The study has nothing specific to do with Apple Watch. They mentioned that they used "a commonly worn commercial wearable device (Apple Watch)" for testing.
The title is factual and not misleading. It seems like you want it changed to remove a fact.
> Any device with continuous heart rate monitoring will do.
This is something you assert, is definitely not a conclusion of the study, and is misleading.
Considering the proliferation of fake pulse-oximeters due the pandemic, it might be easy to go for Apple Watch if one can afford it and have access to it.
I got a FDI approved pulse-oximeter accidentally from Amazon(US)[1] after realizing almost every pulse-oximeters sold online where I live were making false claims[1].
Note though that the study does not use (or even mention) the oximeter. They use heart rate variability.
Edit: They probably also used the older model as the study was submitted in November and the Apple Watch with oximeter came out in September, and was seriously supply constrained.
Whatever happened with the MIT cough analyzer? That seemed like really promising tech that just kind of fell off the news feeds and was never mentioned again. I thought they were trying to apply for an FDA approval. Latest info is 47 days old: https://www.bbc.com/news/technology-54780460?fbclid=IwAR1bGR...
This is basically how the WHOOP band works. Anyone with an Apple Watch curious to see can find your tracked HRV in the Health App on your iPhone.
However, looking at heart-rate (HR) itself will be misleading. Big swings in HR or high HR in general usually misinterpreted as having high HRV, when it's often the exact opposite. HRV is a measure of the variability of 1/(inter-beat time). So it's measuring the time-average of really small variations from a set point.
And counterintuitively, HRV is usually highest when the sympathetic (flight or fight) nervous tone and HR is lower. In those situations, the parasympathetic control dominates, and small wiggles caused by "noise on the wire" conducted directly to the heart's pacemaker via the vagus nerve increases HRV. Often times this is when HR is lower, and because higher HR or spikes in HR are often caused by increased sympathetic tone, this leads to a blunting of this "parasympathetic noise".
I track mine on my Apple Watch which records it in the Health App. It's basic graph and I can download it if I want, but I've found tracking day-over-day/week-over-week is a good enough signal. For example, I did dry January this year and practiced meditation every day since and I've seen a step-jump up corresponding to that.
Still have to keep another record of external factors to correlate with it, though. I've heard people finding success with a 3p monitor and the Elite HRV app as well [0].
I've never had the WHOOP, but I use Apple Watch combined with iPhone apps to track heart metrics. I recently discovered AutoSleep and HeartWatch, which both track HRV along with other metrics.
Ineresting. So... Prior to the diagnosis of COVID-19 by nasal PCR, significant changes in heart rate variability (HRV), were observed demonstrating its predictive ability to identify COVID-19 infection.
So in theory watching your heart rate closely you should be able to see it coming?
(I tried to figure out exactly what to watch for and how to measure it, but couldn't figure it out, hopefully someone else can explain if anyone with a watch could do this)
The measure they’re talking about, HRV, is measuring variance of the peak to peak and trough to trough between successive heartbeats. Individual heartbeats are on the order of one or two a second, but are remarkably consistent, so this measurement of variance is usually on the order of milliseconds.
Healthy ranges are, vaguely speaking, above 80ms, with unhealthy ranges being much lower, the idea being that when you’re sick or stressed, your heart rate control falls back to the nervous system since your brain has other things to deal with, making your heartbeat more consistent, which means lower variability. This is generally not great because you want to be able to switch up the spacing between beats on the fly for better efficiency, kind of like switching gears on a car.
All of this is to say that the measurement they’re talking about is pretty tiny and you’re unlikely to be able to measure it accurately manually.
I'm not sure what's being said here is safe or appropriate. If you're doing nothing, your heart rate variance won't likely change. There's also very little that matters about it, just normals, for you.
Every Google Search I've ever done on HRV has turned up that fact: HRV is incredibly personal and personal trends matter the most.
From my own experience, my HRV is low when I'm not doing anything - and why wouldn't it be? I'm not _doing_ anything. I have no need for sudden changes when I'm typing on a keyboard.
Further, you can be as healthy as a marathon runner and have very low HRVs.
"From my own experience, my HRV is low when I'm not doing anything - and why wouldn't it be? I'm not _doing_ anything. I have no need for sudden changes when I'm typing on a keyboard."
It's not about whether you "have a need for sudden changes" or not, but whether the sympathetic or the parasympathetic nervous system has the upper hand. There are many reasons for why it could be one or the other, but they are usually roughly summarized as "physical or mental stress".
For instance, when you sit at your keyboard and respond to someone who is wrong on the Internet, that is likely to cause negative emotions and mental stress. This would increase sympathetic nervous system activity and as a result lower your HRV.
"Further, you can be as healthy as a marathon runner and have very low HRVs."
Marathon runners are not always equally healthy. They also get sick or stressed sometimes, and quite often they are under physical stress due to training a lot. That's why some of them observe HRV to get an indication of their readiness for high intensity training or competitions (e.g. Whoop, Garmin Recovery Advisor).
That means, just sitting at their keyboard their HRV could be high or low, depending e.g. on how well they recovered from a hard training run.
Increased heart rate with decreased SpO2 commonly precedes diagnosis of respiratory infection. Increased heart rate alone is often seen as a warning to run pathology labs before beginning chemotherapy which can destroy neutrophils. Something you don't want to do if the patient is in the beginning stages of an infection.
I'm not a doctor I've just been around this stuff more than I wish to. Frankly I wish the Apple watch was more aggressive in reporting anomalies to the user. Something as simple as "we noticed your heart rate always increases at this place and time" can make the user aware that something is wrong in their life... assuming it's not a gym.
From my reading, many respiratory illnesses can cause these symptoms, but right now if you have a respiratory illness at all, odd are it's COVID-19, so it's worth at least getting a PCR test. And if it's not that, you have a very serious flu or something which probably requires treatment as well.
You are right, I had similar data and same warning as I got for COVID-19 from Oura cca year ago with regular flu. It is now more important because of COVID-19, but it is also very helpful for any similar illness that one may spread.
There are lots, as what you really are measuring are subtle variations in the sympathetic and parasympathetic drives of the heart by the autonomous nerve system. This can get affected by lots of things, a quick googling provided dozens of completely different papers looking it stuff like heart disease itself, but also clinical depression etc.
Oura Ring also ran a study on this (which I was a part of). It is interesting because in 2019 I had a severe cold and the Oura Ring gave me a very low "Readingess" score in the 2 days leading up to it, despite me feeling fine and getting good sleep. I had never had readiness scores so low prior to that or since, so I don't believe it was any sort of confirmation bias.
Quantified health is going to bring about some amazing QOL improvements that we're still only beginning to understand and it is one of the areas I am most bullish about for sure.
Offtopic: what is up with this website? I just clicked the link and noticed that Firefox downloaded the paper's PDF 55 times automatically before I closed the window!
Is there anything actionable for us, today, in the study? I’m not skilled in reading these, I read the first few paragraphs and knew I was over my head.
I wish I could buy a watch that continuously monitored my hydration levels. So many health concerns are directly impacted by hydration.
Edit: My comment meant to recognize that while this tech doesn’t yet exist, it would be helpful as people end up in the hospital all the time for medical problems that are exacerbated by dehydration. Simply saying “drink water when you’re thirsty” is beside the point.
If you're thirsty then drink some water. We have no reliable clinical evidence that drinking extra water when you're not thirsty causes improved health.
So far no one has succeeded in building a reliable, wearable, non-invasive hydration sensor.
A lot of people have conditioned themselves to ignore the signs of thirst or misinterpret them as hunger. Our, largely, idle/still lives probably doesn't help here, as staying at a survivable level (even if not a truly healthy level) of hydration isn't hard if you're not working up a sweat or in environmental conditions that cause you to become dehydrated faster.
There have been a few cyclists who use whoop and it picked up on HRV and they could see that something was not right.
Hopefully HRV gets added to the AWatch soon, if it isn't already. (C'mon Apple, deliver my device!)
It is pretty cool that we can take these measurements and hopfully make data driven decisions. Hopefully in time we can use these measures to be prescritive and be guided to make better decisions.
Yep. It takes a sample every few hours, but you can also get a HRV measurement on demand by running a session in the Breathe app on the watch.
It doesn't tell you anything about HRV in the Breathe app or that it took a new sample for you, but you'll find a new entry in the Health app if you go look right afterward.
I get that it's cool to ridicule Apple, but maybe pick something they've actually demonstrated a tendency to do in the past? In the case of health features on their Watch they've been pretty consistent about just making it available if the Watch has the right sensors (so older watches don't get all the new features because they lack some of the necessary sensors).
This does not surprise me one bit. I have a child and every time he gets sick his heart rate rises. The last time I noticed the heart rate first and called his mom telling her I think he is getting sick. The next day he had a fever.
I am kind of jaded, but despite the interesting approach, I predict that this research will not end up being deployed in time to make any difference. The real world pipeline just takes too long.
I don't think it's specific for apple, they did mention that the participants just had the apple watch, but from the sounds of it, any GPS hear rate monitor would help..
I think Maybe a collective app simular to Strava might be a better place to get a collection of data maybe..
Oura was quite good with warning me that something is not ok and I should take it easy on the morning of the day when I started to have symptoms. I was tested positive later that day. Interestingly, nobody I met days before I was warned was infected, so in my case warning came in right moment.
I was going to mention them. Some members of the US military were given them as part of a study on COVID detection. Basically any wearable sensors covering the right metrics can be used for this. The question, really, was less about "Can wearable health monitoring equipment detect early signs of <disease>?" than "What are the early signs which wearables can detect and use for prediction once we know them?"
This sounds great, I am going through the PDF, there's a LOT there..
But I'm not sure how reliable that it would be just to track the position (possible infection radius to known area) And the heart rate variability.. Its interesting though! Certainly every little helps..
I really recommend the episode: https://www.pbs.org/wnet/amanpour-and-company/video/how-earl...
Here's another article on a hospital implementing the early monitoring program https://www.nny360.com/news/publicservicenews/samaritan-util....
> With the oximeter, Samaritan is asking patients to keep track of its measurements twice a day and to write the readings down. Then Home Health follows up with telehealth visits, either over the phone or via Zoom on days two and four, as well as days seven and 10.